ggbipart, an R package for plotting bipartite networks
This is a series of R functions aimed to plot bipartite networks within the ggplot2 environment. The library relies heavily on code developed by Francois Briatte for the ggnet library.
Bipartite networks are a special type of network where nodes are of two distinct types or sets, so that connections (links) only exist among nodes of the different sets.
As in other types of network, bipartite strucures can be binary (only the presence/absence of the links is mapped) or quantitative (weighted), where the links can have variable importance or weight.
To plot, we start with an adjacency or incidence matrix. I’m using matrices that illustrate ecological interactions among species, such as the mutualisttic interactions of animal pollinators and plant flowers. The two sets (modes) of these bipartite netwroks are animals (pollinators) ans plants species.
From any adjacency matrix we can get a network object or an igraph object for plotting and analysis.
Required packages
require(ggplot2)
require(network)
require(igraph)
require(sna)
require(ggnet)
require(ggrepel)
require(ergm)
require(intergraph)
require(RColorBrewer)Here I plot bipartite networks from their adjacency matrices, i.e., the two-mode networks. The standard way to input an adjacency matrix is from a .txt or a.csv` file.
Initializing bipartite webs as network objects
# Read data matrices.
# Read a network
# Creating the objects. Example input from the clipboard.
#
# Where data.txt has a weighted adjacency matrix, e.g.,:
Aa Ab Ac Ba Bb Bc Bd Ca Cb Cc Da
P1 139 60 9 23 4 104 5 3 5 2 1
P2 184 26 6 10 3 6 17 11 3 1 0
P3 131 74 33 36 13 19 0 9 0 1 0
P4 87 40 38 21 13 0 9 0 1 0 0
P5 100 42 17 12 4 1 1 0 0 0 0
P6 21 15 7 0 4 1 0 0 0 0 0
P7 42 16 3 0 2 0 0 0 0 0 0
P8 31 8 4 5 1 0 0 1 0 0 0
P9 46 8 2 0 3 0 0 0 0 0 0
P10 86 0 0 12 0 2 2 0 0 0 0
P11 23 8 1 5 0 0 0 0 0 0 0
P12 13 6 1 1 0 1 0 0 0 0 0
P13 5 0 1 0 0 0 0 0 0 0 0
P14 9 1 0 0 0 0 0 0 0 0 0
# Use this to copy from the clipboard, after select/copy the above block.
mymat <- read.table(pipe("pbpaste"), header= T, sep= "\t", row.names= 1)The adjacency matrix is just read from the clipboard as a tab-separated file with header names, and the first column is taken as the row names.
This next example (from F. Briatte code) initializes a dataframe:
# A weighted adjacency matrix. Rows are animal species; columns are plant
# species.
bip= data.frame(P1= c(1, 12, 6, 0),
P2= c(1, 0, 4, 0),
P3= c(1, 7, 3, 12),
row.names= letters[1:4])These are two example datasets of well-sampled plant-frugivore interaction networks from S Spain, read in the usual way.
# The Nava de las Correhuelas dataset.
nch<- read.table("./data-raw/w96_adj_fru.csv",
header=T, sep=",", row.names=1,
dec=".", na.strings="NA")
## Node attributes
nch_attr<- read.table("./data-raw/w96_node_attributes.csv",
header=T, sep=",",
dec=".", na.strings="NA")
# The Hato Raton dataset.
hr<- read.table("./data-raw/w97_adj_fru.csv",
header=T, sep=",", row.names=1,
dec=".", na.strings="NA")
## Node attributes
hr_attr<- read.table("./data-raw/w97_node_attributes.csv",
header=T, sep=",",
dec=".", na.strings="NA")Node attributes include different variables characterizing each individual node. These values can later be passed to the bip_ggnet funtions to modify graph properties of nodes.
glimpse(nch_attr)## Observations: 62
## Variables: 26
## $ class <fctr> Magnoliopsida, Liliopsida, Magnoliopsida, Magnoliopsi...
## $ order <fctr> Rosales, Arales, Ranunculales, Rosales, Rosales, Euph...
## $ family <fctr> Rosaceae, Araceae, Berberidaceae, Rosaceae, Rosaceae,...
## $ genus <fctr> Amelanchier, Arum, Berberis, Cotoneaster, Crataegus, ...
## $ species <fctr> Amelanchier.ovalis, Arum.italicum, Berberis.vulgaris,...
## $ spcode <fctr> Amova, Arita, Bevul, Cogra, Crmon, Dalau, Hehel, Juco...
## $ w <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ g <dbl> 7.3, 9.2, 5.5, 7.0, 9.3, 6.4, 8.4, 8.6, 12.9, 11.5, 8....
## $ f <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ a <dbl> 0.00010, 0.66000, 14.52000, 0.14500, 0.84500, 0.00011,...
## $ geog <int> 4, 7, 14, 6, 14, 10, 6, 13, 4, 8, 8, 9, 4, 5, 3, 5, 5,...
## $ ph <dbl> 1.0, 0.5, 3.0, 1.0, 6.0, 1.0, 3.0, 10.0, 2.0, 7.0, 9.0...
## $ pp <dbl> 0.3333, 0.2121, 0.2424, 0.3030, 0.1515, 0.2727, 0.2424...
## $ ps <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ pa <dbl> 0.1515, 0.1515, 0.1212, 0.0909, 0.1515, 0.2424, 0.0606...
## $ pu <dbl> 0.4242, 0.5758, 0.2121, 0.5455, 0.0303, 0.3636, 0.4545...
## $ hab <fctr> climax, climax, climax, climax, climax, climax, clima...
## $ diet <fctr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N...
## $ forag <fctr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N...
## $ noct <int> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ nativ <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
## $ frlen <dbl> 7.0, 11.1, 10.4, 7.5, 12.1, 9.4, 7.2, 8.2, 11.3, 9.9, ...
## $ fleshy <fctr> Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes...
## $ seedm <dbl> 0.003, 0.038, 0.020, 0.025, 0.124, 0.022, 0.016, 0.011...
## $ height <dbl> 1.9, 0.3, 1.2, 1.3, 4.0, 1.2, 3.0, 0.6, 2.8, 0.6, 0.6,...
## $ grwform <fctr> shrub, herb, shrub, shrub, tree, shrub, liana, shrub,...glimpse(hr_attr)## Observations: 33
## Variables: 26
## $ class <fctr> Aves, Aves, Aves, Aves, Aves, Aves, Aves, Aves, Aves,...
## $ order <fctr> Passeriformes, Passeriformes, Passeriformes, Passerif...
## $ family <fctr> Muscicapidae, Muscicapidae, Muscicapidae, Muscicapida...
## $ genus <fctr> Sylvia, Sylvia, Erithacus, Sylvia, Turdus, Turdus, Sy...
## $ species <fctr> Sylvia.atricapilla, Sylvia.borin, Erithacus.rubecula,...
## $ spcode <fctr> Syatr, Sybor, Errub, Symel, Tumer, Tuphi, Sycom, Syho...
## $ w <dbl> 17.5, 18.7, 15.8, 11.5, 84.0, 65.8, 16.9, 25.0, 10.9, ...
## $ g <dbl> 8.0, 8.0, 8.5, 6.6, 13.7, 12.2, 7.3, 8.3, 5.7, 7.7, 8....
## $ f <dbl> 0.861, 0.924, 0.659, 0.739, 0.889, 0.989, 0.962, 0.660...
## $ a <dbl> 4.692, 2.058, 16.150, 15.908, 11.842, 1.325, 0.650, 0....
## $ geog <dbl> 14, 6, 13, 10, 14, 14, 12, 2, 6, 12, 7, 6, 4, 2, 4, 10...
## $ ph <dbl> 7, 1, 7, 12, 12, 3, 1, 2, 1, 3, 6, 12, 6, 12, 7, 3, 5,...
## $ pp <dbl> 0.063, 0.250, 0.125, 0.000, 0.000, 0.438, 0.375, 0.375...
## $ ps <dbl> 0.000, 0.000, 0.188, 0.125, 0.000, 0.063, 0.063, 0.000...
## $ pa <dbl> 0.000, 0.000, 0.000, 0.000, 0.125, 0.063, 0.000, 0.000...
## $ pu <dbl> 0.000, 0.000, 0.000, 0.000, 0.000, 0.125, 0.000, 0.000...
## $ hab <fctr> forest, forest, forest, other, forest, forest, forest...
## $ diet <fctr> fruit, fruit, fruit, fruit, fruit, fruit, fruit, frui...
## $ forag <fctr> solitary, solitary, solitary, solitary, solitary, soc...
## $ noct <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, NA,...
## $ nativ <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
## $ frlen <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ fleshy <fctr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N...
## $ seedm <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ height <dbl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ grwform <fctr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N...Here I use the function bip_init_network to initialize a few examples of bipartite networks. The function returns a network object. The equivelent fucntion bip_init_igraph returns an igraph (graph) object.
source("./R/bip_init_network.R")
source("./R/bip_init_igraph.R")
# test.net<- bip_init_network(mymat)
# test.ig<- bip_init_igraph(mymat)
nch.net<- bip_init_network(nch)
nch.ig<- bip_init_igraph(nch)
hr.net<- bip_init_network(hr)
hr.ig<- bip_init_igraph(hr)Functions to plot network graphs
The ggbipart package includes the following functions:
bip_qtplot
Simple function to plot a weighted bipartite network in network object. The input is just the weighted (quantitative) adjacency matrix of a two-mode network. Uses objects of type network.
bip_railway
Simple function to plot a bipartite network with the classic layout of two parallel sets of nodes (as in the bipartite package).
bip_ggnet
Plotting bipartite networks from the adjacency matrix of a two-mode network. It uses the ggnet package extensively.
bip_igplot
Function to plot a weighted bipartite network in igraph. Its inputs are the adjacency matrix, to get the dimensions from, and the igraph object corresponding to the matrix.
# Sourcing required functions and initializing the net objects.
source("./R/bip_igplot.R")
source("./R/bip_qtplot.R")
source("./R/bip_railway.R")
source("./R/vectorize.R")
source("./R/bip_edgewt.R")Using ggnet for bipartite graphs
Bipartite networks with ggnet2 using the ggplot2 framework
We first initialize the bipartite network; then uses the adjacency matrix to compute scaled weighted edges with function edgewt. This is based in code from Francois Briatte, using package ggnet, function ggnet2.
ggnet2 automatically detects two-mode graphs from their bipartite network attribute. To simplify the plotting of each mode, it understands arguments of the form [alpha, color, shape, size] = “mode”, which will mark the primary mode as "actor" and the secondary mode as "event".
# Plotting bipartite networks from adjacency matrix of two-mode network.
# Using ggnet.
# DATE: 15Jul2013. Updated to ggnet 11/11/2016.
#
require(ggnet)
require(ggrepel)## Loading required package: ggrepel#
# test.net<- bip_init_network(mymat)
mynet<- nch.net
if(!is.network(nch.net)) stop("Must first initialize the network; use 'bip_init_network.R'.")Node colors
Let’s now assign a vertex attribute called phono, which indicates whether the name of the vertex is a vowel or a consonant:
This attribute can be passed to ggnet2 to indicate that the nodes belong to a group. All the user has to do is to pass the name of the vertex attribute to the color argument, which will find it in the list of vertex attributes and use it to map the colors of the nodes:
# Attributes for animals and plants.
names(nch_attr)
# Load in the preferred attributes
# We need to convert factors to chars in order to assign the attributes.
Fam<- as.character(nch_attr$family)
nch.net %v% "Fam"= Fam
Gen<- as.character(nch_attr$genus)
nch.net %v% "Gen"= Gen
nch.net %v% "BM"= nch_attr$w
nch.net %v% "FrDiam"= nch_attr$gA simple graph
To use the mode of the nodes as the basis for their colors, all the user has to do is to pass the color= “mode” argument, and then to style the “actor” and “event” values:
Unweighted
# Plot layout coordinates for railway networkplot. Input is the
# adjacency matrix.
#
mymat <- read.delim("./data-raw/data.txt", row.names=1) # Not run.
bip_railway(mymat, label=T)## Loading required package: scales
# Detect and color the mode
col= c("P"= "grey80", "A"= "gold2")
ggnet2(nch.net,
color= "mode", palette= col, shape= "mode",
label= TRUE, label.size=3,
layout.exp= 0.25) # Using layout.exp because of long labels
Weighted
# Label weights directly into edges
bip= data.frame(P1= c(1, 12, 6, 0),
P2= c(1, 0, 4, 0),
P3= c(1, 7, 3, 12),
row.names= letters[1:4])
col= c("actor"= "grey80", "event"= "gold2")
bip.net<- network(bip,
matrix.type = "bipartite",
ignore.eval = FALSE,
names.eval = "weights")
ggnet2(bip.net,
color= "mode", palette = col,
label= TRUE, shape= "mode",
edge.label= "weights")## Warning in guide_merge.legend(init, x[[i]]): Duplicated override.aes is
## ignored.
# Set colors for each mode
col= c("P"= "#FC9272", "A"= "#9ECAE1")
pp<- ggnet2(hr.net, size=7,
shape= "mode", #label= T,
color= "mode", palette= col,
edge.size= bip_edgewt(hr, 5),
edge.alpha= 0.25,
layout.exp= 0.25) +
geom_text(aes(label= network.vertex.names(hr.net)),
color= "black", size= 3)
pp
Antother alternative is to pass the node colors directly, with a vector of node colors that has exactly the same length as the number of nodes in the network:
pp1<- ggnet2(nch.net, size= 6, color= "mode", label= T, label.size=3,
palette= c("P"= "indianred3", "A"= "steelblue4"))
pp1
pp2<- ggnet2(nch.net, size= 0,
shape= "mode",
palette= "Set1",
color= "mode",
edge.size= bip_edgewt(nch, 7.5),
edge.alpha= 0.25,
layout.exp= 0.25) +
geom_point(aes(color= color), size= 10, color= "white") +
geom_point(aes(color= color), size= 10, alpha= 0.5) +
geom_point(aes(color= color), size= 8) +
geom_text(aes(label= network.vertex.names(nch.net)),
color= "black", size= 3.5) + # check_overlap= TRUE
guides(color= FALSE)
pp2
nums<- as.vector(c(1:sum(dim(nch))))
pp3<- ggnet2(nch.net, size= 0,
shape= "mode",
palette= "Set1",
color= "mode",
edge.size= bip_edgewt(nch, 7.5),
edge.alpha= 0.5,
layout.exp = 0.25) +
geom_point(aes(color= color), size= 10, color= "white") +
geom_point(aes(color= color), size= 10, alpha= 0.5) +
geom_point(aes(color= color), size= 8) +
geom_text(aes(label= nums),
color= "white", size= 3.5, fontface="bold") +
guides(color= FALSE)
pp3
pp4<- ggnet2(nch.net, size= 7,
shape= "mode",
palette= "Set1",
color= "mode",
edge.size= bip_edgewt(nch, 7.5),
edge.alpha= 0.25,
layout.exp = 0.25) +
guides(color= FALSE) +
# This is how the check overlap option in ggnet2 goes.
# geom_text(aes(label= network.vertex.names(nch.net)),
# color= "black", size= 3.5) + # check_overlap= TRUE
# This uses ggrepel.
geom_text_repel(aes(label= network.vertex.names(nch.net)),
color= "black", size= 3.5)
pp4
sessionInfo()## R version 3.3.2 (2016-10-31)
## Platform: x86_64-apple-darwin13.4.0 (64-bit)
## Running under: macOS Sierra 10.12.4
##
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
##
## attached base packages:
## [1] stats grDevices graphics datasets utils methods base
##
## other attached packages:
## [1] scales_0.4.1 ggrepel_0.6.9 dplyr_0.5.0
## [4] RColorBrewer_1.1-2 taxize_0.8.4 lme4_1.1-12
## [7] Matrix_1.2-7.1 knitr_1.14 intergraph_2.0-2
## [10] popgraph_1.4 igraph_1.0.1 ggnetwork_0.5.1
## [13] ggnet_0.1.0 ggplot2_2.2.1 markdown_0.7.7
## [16] downloader_0.4 bipartite_2.07 sna_2.4
## [19] network_1.13.0 statnet.common_3.3.0 vegan_2.4-0
## [22] lattice_0.20-34 permute_0.9-0 ape_4.1
##
## loaded via a namespace (and not attached):
## [1] Rcpp_0.12.9 assertthat_0.1 rprojroot_1.0-2
## [4] digest_0.6.12 foreach_1.4.3 R6_2.2.0
## [7] plyr_1.8.4 evaluate_0.9 spam_1.3-0
## [10] httr_1.2.1 lazyeval_0.2.0 minqa_1.2.4
## [13] data.table_1.10.0 nloptr_1.0.4 rmarkdown_1.3
## [16] splines_3.3.2 stringr_1.2.0 munsell_0.4.3
## [19] mgcv_1.8-15 htmltools_0.3.5 tibble_1.2
## [22] lpSolve_5.6.13 codetools_0.2-15 reshape_0.8.6
## [25] MASS_7.3-45 grid_3.3.2 nlme_3.1-128
## [28] jsonlite_1.3 gtable_0.2.0 DBI_0.4-1
## [31] magrittr_1.5 formatR_1.4 stringi_1.1.2
## [34] reshape2_1.4.2 sp_1.2-4 xml2_1.1.1
## [37] iterators_1.0.8 tools_3.3.2 bold_0.4.0
## [40] maps_3.1.0 fields_8.4-1 parallel_3.3.2
## [43] colorspace_1.2-6 cluster_2.0.5 sampling_2.8